Matchings and Copeland's Method

Given a graph $G = (V,E)$ where every vertex has weak preferences over its neighbors, we consider the problem of computing an optimal matching as per agent preferences. The classical notion of optimality in this setting is stability. However stable matchings, and more generally, popular matchings need not exist when $G$ is non-bipartite. Unlike popular matchings, Copeland winners always exist in any voting instance -- we study the complexity of computing a matching that is a Copeland winner and show there is no polynomial-time algorithm for this problem unless $\mathsf{P} = \mathsf{NP}$. We introduce a relaxation of both popular matchings and Copeland winners called semi-Copeland winners. These are matchings with Copeland score at least $\mu/2$, where $\mu$ is the total number of matchings in $G$; the maximum possible Copeland score is $(\mu-1/2)$. We show a fully polynomial-time randomized approximation scheme to compute a matching with Copeland score at least $\mu/2\cdot(1-\varepsilon)$ for any $\varepsilon > 0$.